OSX vs Windows in terms of sample efficiency?
-
- KVRist
- Topic Starter
- 45 posts since 14 Jan, 2013 from Boston MA
Hello all, I have a technical question regarding DSP sampler efficiency across platforms.
I'll start by coming right out and saying this post is made with a bit of an axe to grind...
I recently purchased a very pricy sample library from a developer who will remain nameless. (It's not NI, Kontakt has never been an issue in my setup. Vienna runs great as well...)
Anyway here's the details…
The library only works in its own proprietary player available as both Win VST and OSX AU.
1st axe to grind:
Performance of this sample engine on a 12 core Mac Pro is dreadful: 15 GB in this sample engine took 40-50 minutes to load into RAM. Kontakt on the other hand takes 7-8 minutes to load a 40 GB into memory.
(my specs are below.)
2nd axe to grind:
There have been ongoing claims from this developer for over 4 years that:
Windows users have no issues and ultimately there's a "memory issue" in OSX that is the source of these slow load times (and glitchy performance.)
Even users with heavily spec'd machines such as myself are subjected to this same seemingly "pass the buck" explanation of "Mac OS has memory issues that 'said' sample library falls victim to."
So my questions are this.
Is there a dramatically different memory architecture in OSX that would explain such a dramatic difference in performance between Kontakt (as well as Vienna Instruments) vs this sample engine?
E.G. - 2.5 times as much data into RAM at 1/6 the time.
Or does this sound like the work of shoddy coding?
My machine is as follows: 2012 12-core Mac Pro, 60 GB RAM. 2 PCIe SSDs (1 OWC Accelsior E2 SATA IIIx2, 1 Samsung EVO on a PCIe adapter to skirt the Mac Pro SATA II issue and one Crucial M500 on a SATA II Bus. The samples from this library are only coming from the SATA III and IIIx2 drive. Kontakt is the SATA II drive.)
I already have my (assumingly obvious) hunch…
I'll start by coming right out and saying this post is made with a bit of an axe to grind...
I recently purchased a very pricy sample library from a developer who will remain nameless. (It's not NI, Kontakt has never been an issue in my setup. Vienna runs great as well...)
Anyway here's the details…
The library only works in its own proprietary player available as both Win VST and OSX AU.
1st axe to grind:
Performance of this sample engine on a 12 core Mac Pro is dreadful: 15 GB in this sample engine took 40-50 minutes to load into RAM. Kontakt on the other hand takes 7-8 minutes to load a 40 GB into memory.
(my specs are below.)
2nd axe to grind:
There have been ongoing claims from this developer for over 4 years that:
Windows users have no issues and ultimately there's a "memory issue" in OSX that is the source of these slow load times (and glitchy performance.)
Even users with heavily spec'd machines such as myself are subjected to this same seemingly "pass the buck" explanation of "Mac OS has memory issues that 'said' sample library falls victim to."
So my questions are this.
Is there a dramatically different memory architecture in OSX that would explain such a dramatic difference in performance between Kontakt (as well as Vienna Instruments) vs this sample engine?
E.G. - 2.5 times as much data into RAM at 1/6 the time.
Or does this sound like the work of shoddy coding?
My machine is as follows: 2012 12-core Mac Pro, 60 GB RAM. 2 PCIe SSDs (1 OWC Accelsior E2 SATA IIIx2, 1 Samsung EVO on a PCIe adapter to skirt the Mac Pro SATA II issue and one Crucial M500 on a SATA II Bus. The samples from this library are only coming from the SATA III and IIIx2 drive. Kontakt is the SATA II drive.)
I already have my (assumingly obvious) hunch…
-
- KVRist
- Topic Starter
- 45 posts since 14 Jan, 2013 from Boston MA
Ahem… Um, no comment
-
- KVRian
- 1265 posts since 9 Sep, 2005 from Oulu, Finland
Will they deny you all support if you "badmouth" their products publicly? Or what other reason there could be not to just directly name the company and product? It would be important information to know exactly what products have problems and to know what "interesting" explanations their developers have for the problems...dnbhallifax wrote:Ahem… Um, no comment
-
- KVRian
- 573 posts since 1 Jan, 2013 from Denmark
For just loading data from disk to memory there should be virtually no performance difference, i have never experienced anything like that. The difference is quite possibly in their coding and managing of loading samples.
- KVRAF
- 23101 posts since 7 Jan, 2009 from Croatia
PLAY does perform worse on Mac than Windows. I'm also suspecting shoddy coding.
-
- KVRAF
- 6323 posts since 30 Dec, 2004 from London uk
There was a thread on this before re OSX vs Windows. Its not simple but It turns out to be a coding problem :
http://www.kvraudio.com/forum/viewtopic ... 3&t=411000
http://www.kvraudio.com/forum/viewtopic ... 3&t=411000
-
- KVRist
- Topic Starter
- 45 posts since 14 Jan, 2013 from Boston MA
No I'm not worried about that, I was hoping to get an unbiased answer. That being said it's obviously clear I'm referring to Play… Even if I were worried it wouldn't matter. They won't acknowledge this kind of stuff in support and once you buy the license you're stuck with it for life like a bad case of herpes.Xenakios wrote: Will they deny you all support if you "badmouth" their products publicly? Or what other reason there could be not to just directly name the company and product? It would be important information to know exactly what products have problems and to know what "interesting" explanations their developers have for the problems...
And definitely, I'm sure whatever sample encryption or whatever's going on plays a role, but that kind of difference seems excessive.
Windows users do say they have relatively normal load times and it's more stable, but I can't think of any other piece of software with that huge of a performance gap between platforms, so my hunch is shoddy code that was never optimized for OSX.
I've only been using it for a month or two so unfortunately I missed the memo…
-
- KVRist
- Topic Starter
- 45 posts since 14 Jan, 2013 from Boston MA
Thanks for the link to the other thread btw. Pretty fascinating.
Logic certainly does have its quirks in terms of core distribution. Thankfully VEP lets you get around them...
Logic certainly does have its quirks in terms of core distribution. Thankfully VEP lets you get around them...